Sign inSign up

rocm/primus:v26.9

Manifest digest

sha256:921604bee3a7289e95f37db5a4d990dc530f64bdbcabd80efd952aef98647d27

OS/ARCH

linux/amd64

Compressed size

15.59 GB

Last pushed

about 19 hours by rocm

Type

Image

Manifest digest

sha256:921604bee3a7289e95f37db5a4d990dc530f64bdbcabd80efd952aef98647d27

Image Layers

Image Layers

1ARG RELEASE0 B
2ARG LAUNCHPAD_BUILD_ARCH0 B
3LABEL org.opencontainers.image.version=24.040 B
4ADD file ... in / 29.21 MB
5CMD ["/bin/bash"]0 B
6WORKDIR /workspace/99 B
7ENV MAX_JOBS=1280 B
8ENV PYTORCH_ROCM_ARCH=gfx942;gfx9500 B
9ENV ROCM_AMDGPU_TARGETS=gfx942,gfx9500 B
10ENV GPU_ARCHS=gfx942;gfx9500 B
11ENV MLPERF_RUNTIME_SERIES=v26.50 B
12ENV DEBIAN_FRONTEND=noninteractive0 B
13RUN /bin/sh -c apt update236 MB
14COPY /uv /usr/local/bin/uv # buildkit20.96 MB
15RUN /bin/sh -c uv venv9.65 KB
16ENV PATH=/opt/venv/bin:/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin0 B
17ENV VIRTUAL_ENV=/opt/venv0 B
18COPY /dist/pip-*.whl /tmp/ # buildkit1.71 MB
19RUN /bin/sh -c uv pip29.09 MB
20RUN /bin/sh -c apt update95 MB
21ENV HSA_ENABLE_SCRATCH_ASYNC_RECLAIM=00 B
22ENV HSA_NO_SCRATCH_RECLAIM=10 B
23ENV GPU_USE_DEVICE_QUEUE=10 B
24ENV DEBUG_CLR_AQL_DEV_QUEUE=10 B
25ENV TORCH_DIST_USE_NCCL2=00 B
26ENV THE_ROCK_VERSION=10.2.0a202610200 B
27RUN /bin/sh -c uv pip4.19 GB
28ARG PYTORCH_VERSION=2.16.0a0+rocm10.2.0a202610200 B
29RUN |1 PYTORCH_VERSION=2.16.0a0+rocm10.2.0a20261020 /bin/sh -c60.48 MB
30RUN |1 PYTORCH_VERSION=2.16.0a0+rocm10.2.0a20261020 /bin/sh -c1.01 GB
31RUN |1 PYTORCH_VERSION=2.16.0a0+rocm10.2.0a20261020 /bin/sh -c9.3 KB
32RUN |1 PYTORCH_VERSION=2.16.0a0+rocm10.2.0a20261020 /bin/sh -c38.58 KB
33RUN |1 PYTORCH_VERSION=2.16.0a0+rocm10.2.0a20261020 /bin/sh -c4.19 GB
34ENV ROCM_PATH=/opt/venv/lib/python3.12/site-packages/_rocm_sdk_devel0 B
35ENV ROCM_HOME=/opt/venv/lib/python3.12/site-packages/_rocm_sdk_devel0 B
36ENV HIP_PLATFORM=amd0 B
37ENV HIP_PATH=/opt/venv/lib/python3.12/site-packages/_rocm_sdk_devel0 B
38ENV HIP_CLANG_PATH=/opt/venv/lib/python3.12/site-packages/_rocm_sdk_devel/llvm/bin0 B
39ENV HIP_INCLUDE_PATH=/opt/venv/lib/python3.12/site-packages/_rocm_sdk_devel/include0 B
40ENV HIP_LIB_PATH=/opt/venv/lib/python3.12/site-packages/_rocm_sdk_devel/lib0 B
41ENV HIP_DEVICE_LIB_PATH=/opt/venv/lib/python3.12/site-packages/_rocm_sdk_devel/lib/llvm/amdgcn/bitcode0 B
42ENV PATH=/opt/venv/lib/python3.12/site-packages/_rocm_sdk_devel/bin:/opt/venv/lib/python3.12/site-packages/_rocm_sdk_devel/llvm/bin:/opt/venv/bin:/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin0 B
43ENV LD_LIBRARY_PATH=/opt/venv/lib/python3.12/site-packages/_rocm_sdk_devel/lib:/opt/venv/lib/python3.12/site-packages/_rocm_sdk_devel/lib:/opt/venv/lib/python3.12/site-packages/_rocm_sdk_devel/lib64:/opt/venv/lib/python3.12/site-packages/_rocm_sdk_devel/llvm/lib:/opt/venv/lib/python3.12/site-packages/_rocm_sdk_devel/lib/host-math/lib:/opt/venv/lib/python3.12/site-packages/_rocm_sdk_devel/lib/rocm_sysdeps/lib0 B
44ENV LIBRARY_PATH=/opt/venv/lib/python3.12/site-packages/_rocm_sdk_devel/lib:/opt/venv/lib/python3.12/site-packages/_rocm_sdk_devel/lib:/opt/venv/lib/python3.12/site-packages/_rocm_sdk_devel/lib640 B
45ENV CPATH=/opt/venv/lib/python3.12/site-packages/_rocm_sdk_devel/include0 B
46ENV PKG_CONFIG_PATH=/opt/venv/lib/python3.12/site-packages/_rocm_sdk_devel/lib/pkgconfig0 B
47RUN |1 PYTORCH_VERSION=2.16.0a0+rocm10.2.0a20261020 /bin/sh -c28.15 KB
48ARG BWD_ATTN_ASM_REPO=https://github.com/mawad-amd/bwd-attn-asm.git0 B
49ARG BWD_ATTN_ASM_REF=9b9fb6444f3fee388617f62432c3faea740793770 B
50ARG BWD_ATTN_SOURCE_SYMBOL=_ZN5aiter43fmha_bwd_hd64_bf16_causal_a16_rtz_recompileE0 B
51ARG BWD_ATTN_SYMBOL=_ZN5aiter44fmha_bwd_hd64_bf16_causal_a16_rtne_recompileE0 B
52RUN |5 PYTORCH_VERSION=2.16.0a0+rocm10.2.0a20261020 BWD_ATTN_ASM_REPO=https://github.com/mawad-amd/bwd-attn-asm.git BWD_ATTN_ASM_REF=9b9fb6444f3fee388617f62432c3faea740793778.2 KB
53ARG TE_REPO=https://github.com/ROCm/TransformerEngine.git0 B
54ARG TE_VERSION=023639b5168152a339e99d2e06c495dba26cf6cb0 B
55ARG BWD_ATTN_SLOT=bwd_hd64_bf16_causal_a16_rtne.co0 B
56ENV NVTE_USE_HIPBLASLT=10 B
57ENV NVTE_FRAMEWORK=pytorch0 B
58ENV NVTE_ROCM_ARCH=gfx942;gfx9500 B
59ENV NVTE_USE_CAST_TRANSPOSE_TRITON=10 B
60ENV NVTE_CK_IS_V3_ATOMIC_FP32=00 B
61ENV NVTE_CK_USES_BWD_V3=10 B
62ENV NVTE_CK_USES_FWD_V3=10 B
63ENV CK_TILE_FLOAT_TO_BFLOAT16_DEFAULT=20 B
64ENV NVTE_CK_HOW_V3_BF16_CVT=20 B
65ENV NVTE_USE_ROCM=10 B
66ENV NVTE_FUSED_ATTN_AOTRITON=00 B
67ENV NVTE_FUSED_ATTN_CK=10 B
68ENV NVTE_CK_JIT=10 B
69ENV NVTE_CK_FUSED_ATTN_PATH=0 B
70ENV NVTE_AITER_SOURCE_DIR=/workspace/deps/te-aiter0 B
71ENV NVTE_BUILD_MAX_JOBS=1280 B
72ENV CU_NUM=3040 B
73ENV NVTE_SKIP_SUBMODULE_CHECKS_DURING_BUILD=10 B
74ENV CMAKE_BUILD_PARALLEL_LEVEL=1280 B
75RUN |8 PYTORCH_VERSION=2.16.0a0+rocm10.2.0a20261020 BWD_ATTN_ASM_REPO=https://github.com/mawad-amd/bwd-attn-asm.git BWD_ATTN_ASM_REF=9b9fb6444f3fee388617f62432c3faea74079377522.24 MB
76ARG FA_REPO=https://github.com/ROCm/flash-attention.git0 B
77ARG FA_REF=5301a359f59ef8fa10f211618d9f7a69716a88980 B
78RUN |10 PYTORCH_VERSION=2.16.0a0+rocm10.2.0a20261020 BWD_ATTN_ASM_REPO=https://github.com/mawad-amd/bwd-attn-asm.git BWD_ATTN_ASM_REF=9b9fb6444f3fee388617f62432c3faea7407937768.05 MB
79ARG FLUX_REPO=https://github.com/ROCm/AMDiffusionBenchmark.git0 B
80ARG FLUX_BRANCH=26.4_RC_Patch_torchcodec0 B
81COPY pipeline/primus_training/docker/Diffusion_benchmark_patches.patch /workspace/Diffusion_benchmark_patches.patch # buildkit2.81 KB
82COPY pipeline/primus_training/docker/sdxl_patch.patch /workspace/sdxl_patch.patch # buildkit582 B
83COPY pipeline/primus_training/docker/Diffusion_amd_smi.patch /workspace/Diffusion_amd_smi.patch # buildkit766 B
84COPY pipeline/primus_training/docker/siheyuan-dataset-fix.patch /workspace/siheyuan-dataset-fix.patch # buildkit1.04 KB
85COPY pipeline/primus_training/docker/torchcodec_decord_video_shim.patch /workspace/torchcodec_decord_video_shim.patch # buildkit1.89 KB
86RUN |12 PYTORCH_VERSION=2.16.0a0+rocm10.2.0a20261020 BWD_ATTN_ASM_REPO=https://github.com/mawad-amd/bwd-attn-asm.git BWD_ATTN_ASM_REF=9b9fb6444f3fee388617f62432c3faea7407937789.5 MB
87ARG DLRM_REPO=https://github.com/AMD-AGI/DLRMBenchmark.git0 B
88ARG DLRM_BRANCH=main0 B
89RUN |14 PYTORCH_VERSION=2.16.0a0+rocm10.2.0a20261020 BWD_ATTN_ASM_REPO=https://github.com/mawad-amd/bwd-attn-asm.git BWD_ATTN_ASM_REF=9b9fb6444f3fee388617f62432c3faea74079377103.58 KB
90ARG TORCHTUNE_REPO=https://github.com/pytorch/torchtune.git0 B
91ARG TORCHTUNE_BRANCH=b4c98ac2a37f0397d64c22579aed415ce7264db60 B
92RUN |16 PYTORCH_VERSION=2.16.0a0+rocm10.2.0a20261020 BWD_ATTN_ASM_REPO=https://github.com/mawad-amd/bwd-attn-asm.git BWD_ATTN_ASM_REF=9b9fb6444f3fee388617f62432c3faea7407937717.81 MB
93ARG TORCHAO_REPO=https://github.com/pytorch/ao.git0 B
94ARG TORCHAO_BRANCH=e9c7bead90b840b280f97374308255957108ce470 B
95RUN |18 PYTORCH_VERSION=2.16.0a0+rocm10.2.0a20261020 BWD_ATTN_ASM_REPO=https://github.com/mawad-amd/bwd-attn-asm.git BWD_ATTN_ASM_REF=9b9fb6444f3fee388617f62432c3faea740793772.07 MB
96RUN |18 PYTORCH_VERSION=2.16.0a0+rocm10.2.0a20261020 BWD_ATTN_ASM_REPO=https://github.com/mawad-amd/bwd-attn-asm.git BWD_ATTN_ASM_REF=9b9fb6444f3fee388617f62432c3faea74079377173.47 MB
97ARG GROUPED_GEMM_REPO=https://github.com/caaatch22/grouped_gemm.git0 B
98ARG GROUPED_GEMM_BRANCH=rocm0 B
99RUN |20 PYTORCH_VERSION=2.16.0a0+rocm10.2.0a20261020 BWD_ATTN_ASM_REPO=https://github.com/mawad-amd/bwd-attn-asm.git BWD_ATTN_ASM_REF=9b9fb6444f3fee388617f62432c3faea74079377120.39 KB
100ARG CAUSAL_CONV1D_REPO=https://github.com/Dao-AILab/causal-conv1d0 B
101ARG CAUSAL_CONV1D_BRANCH=e940ead2fd962c56854455017541384909ca669f0 B
102ENV CAUSAL_CONV1D_FORCE_BUILD=TRUE0 B
103ENV MAMBA_FORCE_BUILD=TRUE0 B
104ENV HIP_ARCHITECTURES=gfx942,gfx9500 B
105RUN |22 PYTORCH_VERSION=2.16.0a0+rocm10.2.0a20261020 BWD_ATTN_ASM_REPO=https://github.com/mawad-amd/bwd-attn-asm.git BWD_ATTN_ASM_REF=9b9fb6444f3fee388617f62432c3faea740793779.23 MB
106ARG MAMBA_REPO=https://github.com/AndreasKaratzas/mamba.git0 B
107ARG MAMBA_BRANCH=enable-primus-hybrid-models0 B
108RUN |24 PYTORCH_VERSION=2.16.0a0+rocm10.2.0a20261020 BWD_ATTN_ASM_REPO=https://github.com/mawad-amd/bwd-attn-asm.git BWD_ATTN_ASM_REF=9b9fb6444f3fee388617f62432c3faea7407937752.18 MB
109ARG PRIMUS_REPO=https://github.com/AMD-AGI/Primus.git0 B
110ARG PRIMUS_BRANCH=db1a41601833294231daadb3cdbb2933eb805c730 B
111ENV NVTE_FLASH_ATTN=00 B
112ENV NVTE_FUSED_ATTN=10 B
113ENV PRIMUS_FLA_MLA_ATTN=10 B
114RUN |26 PYTORCH_VERSION=2.16.0a0+rocm10.2.0a20261020 BWD_ATTN_ASM_REPO=https://github.com/mawad-amd/bwd-attn-asm.git BWD_ATTN_ASM_REF=9b9fb6444f3fee388617f62432c3faea740793771.7 GB
115ARG AITER_REPO=https://github.com/ROCm/aiter.git0 B
116ARG AITER_COMMIT=778d74d57c765616ee36a0c3295dc446e5027d5f0 B
117RUN |28 PYTORCH_VERSION=2.16.0a0+rocm10.2.0a20261020 BWD_ATTN_ASM_REPO=https://github.com/mawad-amd/bwd-attn-asm.git BWD_ATTN_ASM_REF=9b9fb6444f3fee388617f62432c3faea74079377195.19 MB
118ARG TURBO_REPO=https://github.com/AMD-AGI/Primus-Turbo.git0 B
119ARG TURBO_COMMIT=de8fa6ca7f4986334070d5ee62beee07552f8ad80 B
120ENV HCC_AMDGPU_TARGET=gfx942,gfx9500 B
121ENV PRIMUS_TURBO_FRAMEWORK=PYTORCH0 B
122RUN |30 PYTORCH_VERSION=2.16.0a0+rocm10.2.0a20261020 BWD_ATTN_ASM_REPO=https://github.com/mawad-amd/bwd-attn-asm.git BWD_ATTN_ASM_REF=9b9fb6444f3fee388617f62432c3faea740793771023.33 MB
123RUN |30 PYTORCH_VERSION=2.16.0a0+rocm10.2.0a20261020 BWD_ATTN_ASM_REPO=https://github.com/mawad-amd/bwd-attn-asm.git BWD_ATTN_ASM_REF=9b9fb6444f3fee388617f62432c3faea740793777.04 KB
124RUN |30 PYTORCH_VERSION=2.16.0a0+rocm10.2.0a20261020 BWD_ATTN_ASM_REPO=https://github.com/mawad-amd/bwd-attn-asm.git BWD_ATTN_ASM_REF=9b9fb6444f3fee388617f62432c3faea7407937712.57 MB
125RUN |30 PYTORCH_VERSION=2.16.0a0+rocm10.2.0a20261020 BWD_ATTN_ASM_REPO=https://github.com/mawad-amd/bwd-attn-asm.git BWD_ATTN_ASM_REF=9b9fb6444f3fee388617f62432c3faea7407937754.83 MB
126RUN |30 PYTORCH_VERSION=2.16.0a0+rocm10.2.0a20261020 BWD_ATTN_ASM_REPO=https://github.com/mawad-amd/bwd-attn-asm.git BWD_ATTN_ASM_REF=9b9fb6444f3fee388617f62432c3faea7407937710.84 MB
127WORKDIR /workspace/32 B
128RUN |30 PYTORCH_VERSION=2.16.0a0+rocm10.2.0a20261020 BWD_ATTN_ASM_REPO=https://github.com/mawad-amd/bwd-attn-asm.git BWD_ATTN_ASM_REF=9b9fb6444f3fee388617f62432c3faea740793773.49 MB
129ARG FBGEMM_REPO=https://github.com/pytorch/FBGEMM.git0 B
130ARG FBGEMM_COMMIT=80bd3c077dc41b55cd16ed4dcad15cf7c1c1d76a0 B
131ENV BUILD_ROCM_VERSION=7.150 B
132RUN |32 PYTORCH_VERSION=2.16.0a0+rocm10.2.0a20261020 BWD_ATTN_ASM_REPO=https://github.com/mawad-amd/bwd-attn-asm.git BWD_ATTN_ASM_REF=9b9fb6444f3fee388617f62432c3faea74079377874.19 KB
133RUN |32 PYTORCH_VERSION=2.16.0a0+rocm10.2.0a20261020 BWD_ATTN_ASM_REPO=https://github.com/mawad-amd/bwd-attn-asm.git BWD_ATTN_ASM_REF=9b9fb6444f3fee388617f62432c3faea740793771.71 GB
134RUN |32 PYTORCH_VERSION=2.16.0a0+rocm10.2.0a20261020 BWD_ATTN_ASM_REPO=https://github.com/mawad-amd/bwd-attn-asm.git BWD_ATTN_ASM_REF=9b9fb6444f3fee388617f62432c3faea74079377303.29 KB
135ENV GPU_ARCHS=native0 B
136RUN |32 PYTORCH_VERSION=2.16.0a0+rocm10.2.0a20261020 BWD_ATTN_ASM_REPO=https://github.com/mawad-amd/bwd-attn-asm.git BWD_ATTN_ASM_REF=9b9fb6444f3fee388617f62432c3faea740793778.01 MB
137RUN |32 PYTORCH_VERSION=2.16.0a0+rocm10.2.0a20261020 BWD_ATTN_ASM_REPO=https://github.com/mawad-amd/bwd-attn-asm.git BWD_ATTN_ASM_REF=9b9fb6444f3fee388617f62432c3faea74079377861 B
138RUN |32 PYTORCH_VERSION=2.16.0a0+rocm10.2.0a20261020 BWD_ATTN_ASM_REPO=https://github.com/mawad-amd/bwd-attn-asm.git BWD_ATTN_ASM_REF=9b9fb6444f3fee388617f62432c3faea7407937714.12 MB
139ARG AINIC_BUNDLE_VERSION=1.117.5-a-1960 B
140RUN |33 PYTORCH_VERSION=2.16.0a0+rocm10.2.0a20261020 BWD_ATTN_ASM_REPO=https://github.com/mawad-amd/bwd-attn-asm.git BWD_ATTN_ASM_REF=9b9fb6444f3fee388617f62432c3faea74079377429.17 KB
141RUN |33 PYTORCH_VERSION=2.16.0a0+rocm10.2.0a20261020 BWD_ATTN_ASM_REPO=https://github.com/mawad-amd/bwd-attn-asm.git BWD_ATTN_ASM_REF=9b9fb6444f3fee388617f62432c3faea74079377230 B
142ARG UCX_VERSION=1.18.00 B
143RUN |34 PYTORCH_VERSION=2.16.0a0+rocm10.2.0a20261020 BWD_ATTN_ASM_REPO=https://github.com/mawad-amd/bwd-attn-asm.git BWD_ATTN_ASM_REF=9b9fb6444f3fee388617f62432c3faea7407937786.44 MB
144ENV UCX_INSTALL_DIR=/workspace/ucx-1.18.0/install0 B
145ARG MPI_VERSION=4.1.60 B
146RUN |35 PYTORCH_VERSION=2.16.0a0+rocm10.2.0a20261020 BWD_ATTN_ASM_REPO=https://github.com/mawad-amd/bwd-attn-asm.git BWD_ATTN_ASM_REF=9b9fb6444f3fee388617f62432c3faea740793773.84 MB
147ENV PATH=/opt/openmpi/bin:/opt/venv/lib/python3.12/site-packages/_rocm_sdk_devel/bin:/opt/venv/lib/python3.12/site-packages/_rocm_sdk_devel/llvm/bin:/opt/venv/bin:/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin0 B
148ENV LD_LIBRARY_PATH=/opt/openmpi/lib:/opt/venv/lib/python3.12/site-packages/_rocm_sdk_devel/lib:/opt/venv/lib/python3.12/site-packages/_rocm_sdk_devel/lib:/opt/venv/lib/python3.12/site-packages/_rocm_sdk_devel/lib64:/opt/venv/lib/python3.12/site-packages/_rocm_sdk_devel/llvm/lib:/opt/venv/lib/python3.12/site-packages/_rocm_sdk_devel/lib/host-math/lib:/opt/venv/lib/python3.12/site-packages/_rocm_sdk_devel/lib/rocm_sysdeps/lib0 B
149ARG MLPERF_COMMENT_REPO=https://github.com/AMD-AGI/mlperf-common.git0 B
150ARG MLPERF_COMMENT_COMMIT=7771aeabc8b91402e95cb7a33f447d9d9941361f0 B
151RUN |37 PYTORCH_VERSION=2.16.0a0+rocm10.2.0a20261020 BWD_ATTN_ASM_REPO=https://github.com/mawad-amd/bwd-attn-asm.git BWD_ATTN_ASM_REF=9b9fb6444f3fee388617f62432c3faea7407937725.28 KB
152COPY /wandb-core /opt/venv/lib/python3.12/site-packages/wandb/bin/wandb-core # buildkit15.19 MB
153RUN |37 PYTORCH_VERSION=2.16.0a0+rocm10.2.0a20261020 BWD_ATTN_ASM_REPO=https://github.com/mawad-amd/bwd-attn-asm.git BWD_ATTN_ASM_REF=9b9fb6444f3fee388617f62432c3faea74079377257 B
154RUN |37 PYTORCH_VERSION=2.16.0a0+rocm10.2.0a20261020 BWD_ATTN_ASM_REPO=https://github.com/mawad-amd/bwd-attn-asm.git BWD_ATTN_ASM_REF=9b9fb6444f3fee388617f62432c3faea74079377589 B
155RUN |37 PYTORCH_VERSION=2.16.0a0+rocm10.2.0a20261020 BWD_ATTN_ASM_REPO=https://github.com/mawad-amd/bwd-attn-asm.git BWD_ATTN_ASM_REF=9b9fb6444f3fee388617f62432c3faea7407937732 B
156RUN |37 PYTORCH_VERSION=2.16.0a0+rocm10.2.0a20261020 BWD_ATTN_ASM_REPO=https://github.com/mawad-amd/bwd-attn-asm.git BWD_ATTN_ASM_REF=9b9fb6444f3fee388617f62432c3faea7407937732 B
157RUN |37 PYTORCH_VERSION=2.16.0a0+rocm10.2.0a20261020 BWD_ATTN_ASM_REPO=https://github.com/mawad-amd/bwd-attn-asm.git BWD_ATTN_ASM_REF=9b9fb6444f3fee388617f62432c3faea74079377404 B
158ARG GIT_COMMIT_TAG=8c8b5b3d977cfd4f746ba6cdd96731a894995bc00 B
159ARG DOCKERFILE_PATH=pipeline/primus_training/docker/Dockerfile0 B
160RUN |39 PYTORCH_VERSION=2.16.0a0+rocm10.2.0a20261020 BWD_ATTN_ASM_REPO=https://github.com/mawad-amd/bwd-attn-asm.git BWD_ATTN_ASM_REF=9b9fb6444f3fee388617f62432c3faea7407937721.24 KB
161COPY pipeline/primus_training/docker/Dockerfile /workspace/.manifest/docker-build-recipe.txt # buildkit12.67 KB
162CMD ["/usr/bin/bash"]0 B

Command

 ARG RELEASE